From: Ben Goertzel (ben@goertzel.org)
Date: Sat Jun 08 2002 - 14:14:45 MDT
> (Forwarded here because of some previous discussions about
> whether a seed AI
> team could plausibly work together over the 'Net; I think probably not.)
>
I am well aware of these statistics, and I experienced the phenomenon in
question myself.
At Webmind, we made a mistake when we expanded our office in early 2000;
there was no more room on the 21'st floor so we expanded onto the 16'th
floor, which was also on a different elevator bank. Business & product went
on 16, R&D and IT went on 21. Bad idea. Soon I was the only one going back
and forth every day between both floors. In effect, the company split in 2.
Things were better when Andy (the CEO) and I shared an office with Suzi, one
of the lead AI developers....
On the other hand, the communication between the Brazilian R&D staff and
certain members of US (21'st floor) R&D staff were *extremely* close,
including both planned and unplanned communications, and both formal and
informal interactions.
It seemed to me that
1) when there were "natural distinctions" combined with physical-space
distances separating two groups, the amount of inter-group communication
became VERY VERY SMALL
2) when you had a distributed group working tightly together on a single
project:
a) some people just didn't work out in spite of technical ability and
general people skills, because of somehow not fitting into the "group
communicative vibe"
b) with the right set of people, communication was pretty damn good
Informal, "chance" communications between people on the same team occur in
the form of asides in e-mails, in chat sessions and phone calls that range
widely, etc.
It is *harder* to find a mix of people that can work well as a distributed
team, than to find a mix that can work well as an in-person team. But my
experience is that it is not impossible. I am working right now with a
distributed team and enjoying it very much. Would I rather we were all in
one place? Sure. But would I trade *this* team for any other team, even if
the other team were located all in one place? No again.
Yeah, all the core members of our team know each other in person, and have
met each other in person repeatedly over the years. But there have been a
few recent additions to the team, who none of the other team members has
ever met in person, and who are working out *splendidly* well from my point
of view. DEFINITELY, the process of getting these new distributed team
members up to speed would be TEN TIMES EASIER if they were in the same room
with me and the entire rest of the team. On the other hand, I would take
this handful of new guys in a second over, for instance, THE ENTIRE
CLASSROOM FULL OF STUDENTS from the AI class I taught at UNM this
semester....
In short, I think it is WAY too strong to say that a distributed team "could
not plausibly work" for building a seed AI.
Does having a distributed team pose an extra challenge? Definitely. But I
would rather take that challenge, than take the alternative challenge of
having a team that isn't the best of the best. And the alternative of
bringing the best of the best to one location is only possible if you're
REALLY REALLY well funded. Webmind Inc. was almost there, but not quite,
because, even though we could pay good salaries in New York, some people who
were really great just didn't want to relocate for personal and family
reasons.
-- Ben G
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:39 MDT